A comprehensive index of artificial intelligence and machine-learning research with AI-generated summaries, citation metrics, and direct links to papers and code.
Hmrishav Bandyopadhyay, Xuanchi Ren, Zijian Huang, et al.
Introduces Context-Matched Distillation (CMD), a causal DMD framework that aligns teacher supervision with the information available during autoregressive video generation, improving control adherence and state-of-the-art performance.
Xinyu Zhao, Zhen Tan, Vaishnav Tadiparthi, et al.
SkillComposer formalizes structured skill composition as task-conditioned skill sequence prediction, jointly predicting skill subset, count, and order via a constrained autoregressive decoder, improving coding agent pass rates by up to 23.1pp.
Leach, Adam
This paper provides a comprehensive comparative review of deep generative models, including VAEs, GANs, normalizing flows, energy-based models, and autoregressive models, highlighting their trade-offs and interconnections.
Unknown
This survey provides a comprehensive overview of speculative decoding and related techniques for accelerating autoregressive generation, including draft models and iterative refinement methods.
Shen Nie, Fengqi Zhu, Zebin You, et al.
This paper demonstrates that diffusion models can achieve competitive performance with autoregressive LLMs on core language tasks at scale, challenging the assumption that such capabilities inherently require autoregressive architectures.
Jiaxing Li, Kai Zou, Cindy Zhou, et al.
Proposes joint distillation combining mode-seeking and mode-covering objectives to improve distributional alignment in autoregressive video distillation.
Maarten Blaauw, J. Andrés Christen
Introduces Bacon, a Bayesian autoregressive gamma process for flexible paleoclimate age-depth modeling with robust MCMC sampling.
Siqi Kou, Lanxiang Hu, Zhe He, et al.
Consistency LLMs accelerate LLM inference by refining the model to consistently predict the fixed point from any state, achieving 2.4-3.4x speedup over autoregressive decoding.
Shen Nie, Fengqi Zhu, Zebin You, et al.
LLaDA is a diffusion model trained from scratch for language modeling that matches autoregressive LLMs like LLaMA3 8B across benchmarks and solves the reversal curse.
Unknown
AIM V2 introduces a family of open vision encoders from 300M to 3B parameters, pre-trained with a multimodal autoregressive approach that jointly generates image patches and text tokens using a causal decoder.
Unknown
AIM shows that autoregressive pretraining on images scales effectively with model and data size, achieving strong performance on vision tasks.
Unknown
GPT-NeoX-20B is the largest dense autoregressive LLM with publicly available weights at the time of submission, trained on the Pile dataset.